Kyvvu

E-commerce

Agents that serve your customers, not whoever messages them.

Online shops put agents on customer service, the product catalogue and order operations, where every customer message, review and supplier feed is outside input. Kyvvu makes sure those agents ask before they refund or delete, and never send customer data out.

service-agent · task 7d14kyvvu engine · in-process
  1. step.messagePOSTsmtp/externalblock
Blocked: outbound send after personal data was read in this taskTry it: click the highlighted step to take it off the path

The problem

Every customer is an input.

  1. 01

    Anyone can write to your agent

    Chat messages, emails, reviews and returns forms all reach the agent. Any of them can carry an instruction it wasn't meant to follow.

    prompt injection

  2. 02

    Refunds and credits are money

    A refund, voucher or store credit an agent shouldn't have given is gone, and a trick that works once gets shared.

    refunds/vouchers

  3. 03

    Customer data at scale

    Names, addresses, order histories and contact details for every customer: the data an agent needs to help, and the data that must never leave.

    personal data

In practice

A normal day in the shop.

A refund waits for a person

A customer service agent handles a return and goes to issue the refund straight away.

service-agent · task 1e56kyvvu engine · in-process

Illustrative scenarios. Each one is decided by a single policy from the open policy library, as the engine evaluates it.

How Kyvvu helps

Rules that hold on your busiest day.

  1. 01

    Judges the whole path

    Every action is checked against everything the agent did earlier in the task, and blocked before it runs if the sequence is dangerous.

    allow/warn/block

  2. 02

    Runs inside your environment

    The engine sits in the agent's own process: so there's no proxy or extra network hop, and a check takes under a millisecond. Prompts and data stay with you.

    in-process/no proxy

  3. 03

    Nothing breaks on day one

    Only critical policies block; high, medium and low ones warn. Start a new rule as a warning and make it critical once you trust it.

    shadow/suggest/enforce

  4. 04

    Rules people can read

    Policies are short YAML files in your own git, reviewed and versioned like code. Every verdict in the trace names the rule that decided.

    policies as code

  5. 05

    A record of every step

    Each task is recorded as an ordered trace of its steps with the verdict on each, hash-chained when it reaches the platform.

    logs/traces

  6. 06

    One policy, every agent

    The same rules for plain Python, LangChain, LangGraph, CrewAI and Claude Code agents, in the cloud, on-prem or air-gapped.

    Python/LangChain/CrewAI/Claude Code

Questions

Common questions.

01

Will it slow down our customer service?

No extra network hop: the engine decides in the agent's own process in under a millisecond.

02

Does customer data reach Kyvvu?

No. The engine decides inside the agent's process, so prompts and data stay with you. Traces of steps go to the platform unless you keep them on your own machines.

03

Do we have to rebuild our agents?

No. Kyvvu works with Python, LangChain, LangGraph, CrewAI and Claude Code agents, and with agents in any language through the local engine.

Get started

Every agent has to ASK before it acts.

Start with one agent

See it on your own agents

A 30-minute demo with someone who builds Kyvvu: your agent, your rules, and the policy that decides each step.

For developers

Five minutes, no account

kyvvu try runs a demo agent against the baseline on your machine.

Start free

Want to see other teams and sectors? All use cases.